Remote Sensing systems are exposedto the constantly changing conditions in the open world. The conditions inwhich remote sensing systems operate can change frequently and drastically dueto differences in acquisition (i.e., different atmospheric effects, acquisitionviewpoints, time periods, etc.) and the heterogeneity of the environments underinvestigation (e.g., planetary exploration, sustainable development, UAV anddrone-based data collection). Researchers have been studying challengesinvolved with novelty detection, identification and adaptation to build morereliable and intelligent remote sensing systems.
Robust performance of automatedsystems for open world remote sensing hinges on determining when conditionshave changed enough to jeopardize accurate performance, and how to identifycritical changes so as to appropriately adapt to them. Novelty detection aimsto determine when such changes have occurred without knowledge of what thechanges might be. Lifelong learning aims to support robust performance undernovel conditions by characterizing these changes and continuously adaptingwithout catastrophically forgetting previous knowledge.
- Ryan Brown, Andrew Brna, Jared Cook, Samuel Park and Mario Aguilar-Simon, "Uncertainty-driven Control For A Self-supervised Lifelong Learning Drone", get presentation
- Jian Peng, Dingqi Ye and Lorenzo Bruzzone, "Asymmetric Collaborative Network: Transferable Lifelong Learning For Remote Sensing Images", get presentation
- Surbhi Sharma, Ribana Roscher, Morris Riedel, Shahbaz Memon and Gabriele Cavallaro, "Improving Generalization For Few-shot Remote Sensing Classification With Meta-learning", get presentation
- John E. Vargas-Muñoz, Diego Schibli and Devis Tuia, "Towards Efficient Correction Of Coconut Tree Detection Errors", get presentation
- Subhojit Mandal, Mainak Thakur, Anish C. Turlapaty, Riyaaz Uddien Shaik and Laneve Giovanni, "Application Of Prisma Hyperspectral Data For Pm2.5 Estimation: A Case Study On New Delhi, India", get presentation
- Dawei Du, Christopher Funk and Anthony Hoogs, "Novelty Detection In Remote Sensing Imagery", get presentation
- Matthew D. McLure and David J. Musliner, "A Changepoint Method For Open-world Novelty Detection"
- Hannah Kerner and Jacob Adler, "Guiding Field Exploration On Earth And Mars With Outlier Detection", get presentation
- Tom Stephens†, Isaac Corley, Adrian Gould, Anthony Polakiewicz, David McVicar, Carlos Torres, Rose Colangelo and Mario Aguilar-Simon, "Self-supervised Representation Learning Enhances Broad Area Search In Multi-temporal Satellite Imagery"
- Qingsong Xu, Yilei Shi and Xiaoxiang Zhu1, "Universal Domain Adaptation Without Source Data For Remote Sensing Image Scene Classification"
Katarina Doctor (U.S. Naval Research Laboratory)
Gabriele Cavallaro (Forschungszentrum Jülich)